Image is reconstructed by editing A reconstruction technique used in diffuse optic tomography to improve image quality and obtained images are compared

Objective: Aim of this study is to improve a reconstruction technique which is used in DOT and to determine iteration number of reconstruction techniques automatically. Method: In experiments, Intralipid and Indocyanine green (ICG) was used to make phantom having optical properties of breast tissue. To make inclusion having optical properties of tumor, a transparent balloon was filled with Intralipid and ICG to set absorption coefficients 0.016 cm-1. In this study, scattering coefficients of phantom and inclusion is same, however, absorption coefficients are different thus images base on the differences in the absorption concentration. According to designed system, appropriate reconstruction technique should be decided. It is called as over-determined system because our system consists of 2401 measurement and 2250 voxels. Therefore, it is decided that Truncated Conjugate Gradient (TCG) is most suitable technique for our system. However, TCG has some disadvantages. We used Transpose Free Quasi Minimal Residual (TFQMR) to overcome these problems. Nevertheless, we combined both algorithms as TCG-TFQMR because this method works slowly. In the stage of creating image, an algorithm to determine iteration number was designed. This algorithm based on Contrast to Noise Ratio (CNR) decides to suitable iteration number. Results: To test TCG-TFQMR, we prepared experiment sets. Firstly, simulation images were created with new algorithm after simulation data generated in MATLAB. Later, 3d images were created by TCG and TCG-TFQMR algorithms after in-vitro experiments were done. Images generated by both algorithms were compared to each other in terms of some parameters such as distance, depth and resembling the actual shape. Conclusion: As results obtained by both algorithm compared to each other, the results obtained by TCG-TFQMR are closer to the real and contains less artifact in certain depth. According to our findings, it is observed that TCG-TFQMR algorithm works better in case that TCG fails for over-determined system.